Anyone know how HSpice does their nonlinear resistors?  You can specify
R1 n1 n2 'some_expression'

and it seems to integrate some_expression such that the operating-point resistance gives you back exactly some_expression.

The simplistic approach would be to say
 I(V) = V / some_expression(V)

but in that case, the op-pt conductance would be 1/s_e(V) - V / s_e(V)^2 * d(s_e)/dV and would not give the "correct" resistance.

It's easy to do the simplistic approach in Verilog-A; it's not clear to me how to do the other for a general expression.